DHCP client roaming across scopes doesn't work correctly
jho1670 opened this issue · 3 comments
First, thanks for your work. I rely on this for an internal DNS resolver and it's great.
I tried migrating my MS DHCP server and ran into trouble.
I have two wireless SSIDs, each which tags clients on VLAN 20 and VLAN 40 respectively. A Ruckus/Brocade switch has a VE on each VLAN and the "helper-address" is set relay DHCP traffic the Gravity instance(s). In Gravity, I have two scopes 10.0.20.x and 10.0.40.x.
Using the hardware MAC:
If I connect a client to VLAN 20, it (successfully) obtains a lease like 10.0.20.100/24 (with router/gateway 10.0.20.1). If I then connect this client to VLAN 40, it fails to get a lease on 10.0.40.x.
If I get the client to use a random MAC, this appears to work successfully.
Is Gravity enforcing that the MAC address is unique across the two scopes? Is there a way to not enforce this, if so?
Gravity assumes to be MAC addresses (or rather DHCP client identifiers) to be globally unique, which is not necessarily the best idea; I suppose as a workaround you could set a relatively low TTL?
Makes sense.
I believe my DHCP relay appends additional VLAN hints - is it possible to use those with the provided client identifier?
Woot. Thank you!